Including biotic factors does not improve species distribution models across three trophic levels

Species distribution is broadly influenced by abiotic factors such as elevation and seasonal variations in temperature and precipitation. Further, for specialist herbivores like Manduca sexta and M. quinquemaculata , biotic factors, including host plant availability and the presence of a shared parasitoid Cotesia congregata , may play a critical role in shaping their spatial patterns. Yet, existing predictive models often focus solely on abiotic factors, with the contribution of biotic factors largely unknown.

In this study, we assessed whether abiotic and biotic factors as isolated or combined components of species distribution models (SDM) better predict the distribution of the congeners, M. sexta and M. quinquemaculata , and their shared endoparasitoid , C. congregata. We further evaluated the influence of bioclimate variables and elevation on cultivated and wild host plant distributions of M. sexta and M. quinquemaculata . To ensure robustness of our SDMs, we constructed them using an ensemble of three predictive algorithms: generalized linear model, maximum entropy, and random forest. We also applied Area Under the Curve (AUC) and True Skill Statistic (TSS) metrics to assess how well each model predicted outcomes. Then, to compare the predictive power between models that used isolated and combined abiotic and biotic factors, we analyzed the values obtained from the AUC and TSS metrics using a one-way ANOVA.

Our findings showed that the predictive power of models that used isolated and combined abiotic and biotic factors did not significantly differ. The abiotic-only model identified minimum temperature of the coldest month, elevation, and summer precipitation as the primary determinants of distributions of the Manduca spp. , host plants, and C. congregata, respectively. Both Manduca spp. showed similar responses to minimum temperature of the coldest month and elevation, but response to summer precipitation differed between species. Models accounting for biotic factors only and combination of abiotic and biotic factors revealed cultivated host plants as the most important predictor for the distribution of both the Manduca spp. and C. congregata . Both Manduca spp. And C. congregata showed a high likelihood of occurrence with the presence of cultivated host plants. While this finding is not surprising, the responses of Manduca spp. to predictors like wild host plants varied between biotic-only and abiotic-biotic models. Similarly, the parasitoid’s response to its host, Manduca spp. , also differed between the two models.

We therefore suggest that models incorporating either abiotic or biotic factors, or both, can reliably predict species distribution. However, the direction of species responses whether an increase or decrease in likelihood of occurrence may vary depending on the factors included in the model.